About two weeks ago I sent in a failure report that effected running
SLEPcWrappers but not PETScWrappers. (Sorry I seem to have lost the
original email and replies). Andrew suggested to check the test-suite
(silly me I should have done that first!)
Here are my conclusions based on the results from tests on all supported
PETSc versions and all supported SLEPc versions. Note: This is *only* true
for versions of PETSc and SLEPc that deal.II are compatible with version
6.2.* (and dev of course). That's about 32 configuration possibilities, so
it took some time to get the full picture....
(i) --with-clanguage=C++ is now optional for PETSc (why? what happened?)
(ii) --with-clanguage=C++ does not work for SLEPc >= 3.1 do not use it!
It is in (ii) that the problem occurs; and that is why step-36 dies.
Forget --with-clanguage option and deal.II works very happily with SLEPc.
My current advice is to forget the --with-clanguage=C++ option when
compiling PETSc and SLEPc together; or just completely forget this option
since PETSc seems to be fine without it (is there a performance
difference? I do not know). I do not know yet why SLEPc fails with
this option... a bug report has been filed and Jose E. Roman has confirmed
that this may be a problem; we will work on it and hopefully there will be
a proper fix eventually so we can go back to --with-clanguage=C++ (and
PETSc complex).
In the mean time I am slowly updating the Wiki page based on these
test-suite results and the fact that the wiki page was written for
PETSc versions that are no longer supported by deal.II....
Things will change in the future, but nevertheless, there will be no
revolutions today ;-)
Best,
Toby
-----
Toby D. Young
Assistant Professor
Institute of Fundamental Technological Problems
Polish Academy of Sciences
ul Adolfa Pawinskiego 5b
02-106 Warsaw
Poland
www: http://www.ippt.gov.pl/~tyoung
skype: stenografia
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
